Generic Sketch-Based Retrieval Learned without Drawing a Single Sketch

Training data, models and code coming soon...

Multiple-Kernel Local-Patch Descriptor

Code coming soon...

Diffusion on Region Manifolds

Matlab code to reproduce the results of our CVPR 2017 paper. It implements global and regional diffusion ranking.

Unsupervised fine-tuning of CNN for retrieval

Matlab code to reproduce the results of our ECCV 2016 paper.

R-MAC and Integral Max-pooling Localization

A Matlab/MEX implementation of the methods proposed in our ICLR 2016 paper. It implements MAC, R-MAC and localization with integral max-pooling.

Kernelized local descriptor

A Matlab implementation of the kernelized local descriptor proposed in our ICMR 2015 paper. It implements the extraction of the local descriptor and evaluation on the Brown patch dataset.

Angular modulation

A Matlab implementation of the angular modulation proposed in our ECCV 2014 paper. It implements modulation of VLAD, modulation of monomial embedding, and the trigonometric polynomial for efficient similarity computation of multiple rotations.

Selective match kernels

A Matlab/Mex implementation of the ASMK* variant (good compromise accuracy/speed/memory) is available here. The method is described in this paper.

Hough Pyramid Matching for Large Scale Image Retrieval

Binary code which implements the method described in our paper for large scale image retrieval. See the project page.

Demo: Visual Image Retrieval and Localization (VIRaL)

See the online demo.